{"id":3086,"date":"2016-04-25T16:17:55","date_gmt":"2016-04-25T20:17:55","guid":{"rendered":"http:\/\/cpip.gmu.edu\/?p=3086"},"modified":"2026-02-03T21:11:49","modified_gmt":"2026-02-03T21:11:49","slug":"separating-fact-from-fiction-in-the-notice-and-takedown-debate","status":"publish","type":"post","link":"https:\/\/blogs.uakron.edu\/ualawip\/2016\/04\/25\/separating-fact-from-fiction-in-the-notice-and-takedown-debate\/","title":{"rendered":"[Archived Post] Separating Fact from Fiction in the Notice and Takedown Debate"},"content":{"rendered":"

By Kevin Madigan<\/a> & Devlin Hartline<\/a><\/em><\/p>\n

\"U.S.With the Copyright Office undertaking a new study<\/a> to evaluate the impact and effectiveness of the Section 512<\/a> safe harbor provisions, there\u2019s been much discussion about how well the DMCA\u2019s notice and takedown system is working for copyright owners, service providers, and users. While hearing from a variety of viewpoints can help foster a healthy discussion, it\u2019s important to separate rigorous research efforts from overblown reports that offer incomplete data in support of dubious policy recommendations.<\/p>\n

Falling into the latter category is Notice and Takedown in Everyday Practice<\/em><\/a>, a recently-released study claiming to take an in-depth look at how well the notice and takedown system operates after nearly twenty years in practice. The study has garnered numerous headlines that repeat its conclusion that nearly 30% of all takedown requests are \u201cquestionable\u201d and that echo its suggestions for statutory reforms that invariably disfavor copyright owners. But what the headlines don\u2019t mention is that the study presents only a narrow and misleading assessment of the notice and takedown process that overstates its findings and fails to adequately support its broad policy recommendations.<\/p>\n

Presumably released to coincide with the deadline for submitting comments to the Copyright Office on the state of Section 512, the authors claim to have produced \u201cthe broadest empirical analysis of the DMCA notice and takedown\u201d system to date. They make bold pronouncements about how \u201cthe notice and takedown system . . . meets the goals it was intended to address\u201d and \u201ccontinues to provide an efficient method of enforcement in many circumstances.\u201d But the goals identified by the authors are heavily skewed towards service providers and users at the expense of copyright owners, and the authors include no empirical analysis of whether the notice and takedown system is actually effective<\/em> at combating widespread piracy.<\/p>\n

The study reads more like propaganda than robust empiricism. It should be taken for what it is: A policy piece masquerading as an independent study. The authors\u2019 narrow focus on one sliver of the notice and takedown process<\/em>, with no analysis of the systemic results<\/em>, leads to conclusions and recommendations that completely ignore the central issue of whether Section 512 fosters an online environment that adequately protects the rights of copyright owners. The authors conveniently ignore this part of the DMCA calculus and instead put forth a series of proposals that would systematically make it harder for copyright owners to protect their property rights.<\/p>\n

To its credit, the study acknowledges many of its own limitations. For example, the authors recognize that the \u201cdominance of Google notices in our dataset limits our ability to draw broader conclusions about the notice ecosystem.\u201d Indeed, over 99.992% of the individual requests in the dataset for the takedown study were directed at Google, with 99.8% of that dataset directed at Google Search in particular. Of course, search engines do not include user-generated content\u2014the links Google provides are links that Google itself collects and publishes. There are no third parties to alert about the takedowns since Google is taking down its own content. Likewise, removing links from Google Search does not actually remove the linked-to content from the internet.<\/p>\n

The authors correctly admit that \u201cthe characteristics of these notices cannot be extrapolated to the entire world of notice sending.\u201d A more thorough quantitative study would include data on sites that host user-generated content, like YouTube and Facebook. As it stands, the study gives us some interesting data on one search engine, but even that data is limited to a sample size of 1,826 requests out of 108 million over a six-month period in mid-2013. And it\u2019s not even clear how these samples were randomized since the authors admittedly created \u201ctranches\u201d to ensure the notices collected were \u201cof great substantive interest,\u201d but they provide no details about how<\/em> they created these tranches.<\/p>\n

Despite explicitly acknowledging that the study\u2019s data is not generalizable, the authors nonetheless rely on it to make numerous policy suggestions that would affect the entire notice and takedown system and that would tilt the deck further in favor of infringement and against copyright owners. They even identify some of their suggestions as explicitly reflecting \u201cPublic Knowledge\u2019s suggestion,\u201d which is a far cry from a reasoned academic approach. The authors do note that \u201cany changes should take into account the interests of . . . small- and medium-sized copyright holders,\u201d but this is mere lip service. Their proposals would hurt copyright owners of all shapes and sizes.<\/p>\n

The authors justify their policy proposals by pointing to the \u201cmistaken and abusive takedown demands\u201d that they allegedly uncover in the study. These so-called \u201cquestionable\u201d notices are the supposed proof that the entire notice and takedown system needs fixing. A closer look at these \u201cquestionable\u201d notices shows that they\u2019re not nearly so questionable. The authors claim that 4.2% of the notices surveyed (about 77 notices) are \u201cfundamentally flawed because they targeted content that clearly did not match the identified infringed work.\u201d This figure includes obvious mismatches, where the titles aren\u2019t even the same. But it also includes ambiguous notices, such as where the underlying work does not match the title or where the underlying page changes over time.<\/p>\n

The bulk of the so-called \u201cquestionable\u201d notices comes from those notices that raise \u201cquestions about compliance with the statutory requirements\u201d (15.4%, about 281 notices) or raise \u201cpotential fair use defenses\u201d (7.3%, about 133 notices). As to the statutory requirements issue, the authors argue that these notices make it difficult for Google to locate the material to take down. This claim is severely undercut by the fact that, as they acknowledge in a footnote, Google complies with 97.5% of takedown notices overall. Moreover, it wades into the murky waters of whether copyright owners can send service providers a \u201crepresentative list\u201d of infringing works. Turning to the complaint about potential fair uses, the authors argue that copyright owners are not adequately considering \u201cmashups, remixes, or covers.\u201d But none of these uses are inherently fair, and there\u2019s no reason to think that the notices were sent in bad faith just because someone might<\/em> be able to make a fair use argument.<\/p>\n

The authors claim that their \u201crecommendations for statutory reforms are relatively modest,\u201d but that supposed modesty is absent from their broad list of suggestions. Of course, everything they suggest increases the burdens and liabilities of copyright owners while lowering the burdens and liabilities of users, service providers, and infringers. Having overplayed the data on \u201cquestionable\u201d notices, the authors reveal their true biases. And it\u2019s important to keep in mind that they make these broad suggestions that would affect everyone in the notice and takedown system after explicitly acknowledging that their data \u201ccannot be extrapolated to the entire world of notice sending.\u201d Indeed, the study contains no empirical data on sites that host user-generated content, so there\u2019s nothing whatsoever to support any changes for such sites.<\/p>\n

The study concludes that the increased use of automated systems to identify infringing works online has resulted in the need for better mechanisms to verify the accuracy of takedown requests, including human review. But the data is limited to small surveys with secret questions and a tiny fraction of notices sent to one search engine. The authors offer no analysis of the potential costs of implementing their recommendations, nor do they consider how it might affect the ability of copyright owners to police piracy. Furthermore, data presented later in the study suggests that increased human review might have little effect on the accuracy of takedown notices. Not only do the authors fail to address the larger problem of whether the DMCA adequately addresses online piracy, their suggestions aren\u2019t even likely to address the narrower problem of inaccurate notices that they want to fix.<\/p>\n

Worse still, the study almost completely discards the ability of users to contest mistaken or abusive notices by filing counternotices. This is the solution that\u2019s already built into the DMCA, yet the authors inexplicably dismiss it as ineffective and unused. Apart from providing limited answers from a few unidentified survey respondents, the authors offer no data on the frequency or effectiveness of counternotices. The study repeatedly criticizes the counternotice system as failing to offer \u201cdue process protection\u201d to users, but that belief is grounded in the notion that a user that fails to send a counternotice has somehow been denied the chance. Moreover, it implies a constitutional right that is not at issue when two parties interact in the absence of government action. The same holds true for the authors’ repeated\u2014and mistaken\u2014invocation of \u201cfreedom of expression.\u201d<\/p>\n

More fundamentally, the study ignores the fact that the counternotice system is stacked against copyright owners. A user can simply file a counternotice and have the content in question reposted, and most service providers are willing to repost the content following a counternotice because they\u2019re no longer on the hook should the content turn out to be infringing. The copyright owner, by contrast, then faces the choice of allowing the infringement to continue or filing an expensive lawsuit in federal court. The study makes it sound like users are rendered helpless because counternotices are too onerous, but the reality is that the system leaves copyright owners practically powerless to combat bad faith counternotices.<\/p>\n

Pretty much everyone agrees that the notice and takedown system needs a tune up. The amount of infringing content available online today is immense. This rampant piracy has resulted in an incredible number of takedown notices being sent to service providers by copyright owners each day. Undoubtedly, the notice and takedown system should be updated to address these realities. And to the extent that some are abusing the system, they should be held accountable. But in considering changes to the entire system, we should not be persuaded by biased studies based on limited (and secret) datasets that provide little to no support for their ultimate conclusions and recommendations. While it may make for evocative headlines, it doesn\u2019t make for good policy.<\/p>\n","protected":false},"excerpt":{"rendered":"

By Kevin Madigan & Devlin Hartline With the Copyright Office undertaking a new study to evaluate the impact and effectiveness of the Section 512 safe harbor provisions, there\u2019s been much discussion about how well the DMCA\u2019s notice and takedown system is working for copyright owners, service providers, and users. While hearing from a variety of […]<\/p>\n","protected":false},"author":3627,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[10,29,34,1],"tags":[447,625,816,1028,1330,1500],"class_list":["post-3086","post","type-post","status-publish","format-standard","hentry","category-copyright","category-internet","category-legislation","category-uncategorized","tag-dmca","tag-google","tag-junk-science","tag-notice-and-takedown","tag-section-512","tag-u-s-copyright-office"],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"_links":{"self":[{"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/posts\/3086","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/users\/3627"}],"replies":[{"embeddable":true,"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/comments?post=3086"}],"version-history":[{"count":1,"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/posts\/3086\/revisions"}],"predecessor-version":[{"id":15871,"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/posts\/3086\/revisions\/15871"}],"wp:attachment":[{"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/media?parent=3086"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/categories?post=3086"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blogs.uakron.edu\/ualawip\/wp-json\/wp\/v2\/tags?post=3086"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}